The overall goal of the proposed research is to determine how brain events at the biophysical level in the human brain influence the macroscopic recordings outside the skull. The specific focus is on sleep spindles, the best studied sleep rhythm, for which we will obtain simultaneous EEG, MEG and depth electrode recordings. These empirical data will be analyzed to determine the laminar sources of the currents within the cortex associated with spindles and these results will be matched to predictions from computational models. We will develop three types of related network models to cover a wide range of spatial scales. The working hypotheses are, first, that spindles are generated by multiple loosely-coupled cortical regions through rhythmic activation of thalamocortical afferents, onto superficial cortical layers from the matrix system, and onto middle layers from the core system, and second, that spindle oscillations in restricted thalamocortical domains in the core system are spread via the matrix system. Empirical Specific Aims and Hypotheses 1) Record MEG and EEG simultaneously during sleep spindles. Hypothesis: MEG and EEG will record very different activity patterns at the sensor level and in their inferred sources. Poor correlation will be apparent as inconstant phase and amplitude relations within spindle discharges, as well as only loose correlations as to when spindles occur. 2) Record from intracranial EEG macro-electrodes during sleep spindles, with simultaneous scalp EEG and MEG recordings. Hypothesis: Recordings from different cortical generators will be only loosely coupled with each other, or with scalp EEG, thus supporting the general view suggested by MEG. 3) Record from intracranial microelectrode arrays during sleep spindles. Hypothesis: Neuronal generator currents vary during each spindle between superficial and middle layers. 4) Reconstruct supragranular and infragranular pyramidal cells from association cortex in humans, determine their laminar distribution, and estimate the likely termination zones of core and matrix thalamocortical projections. Hypothesis: Significant anatomical differences will be found between human association cortex and the rodent sensory cortex. Modeling Specific Aims and Hypotheses 1) Construct accurate, realistic EEG/MEG forward solutions based on cortical reconstruction from structural MRI. Hypothesis: Basic parameters of extracranial EEG/MEG can be replicated by modeling multiple thalamocortical domains with varying synchrony between domains. The values of the fit parameters to EEG/MEG will match those inferred from intracranial macroelectrode recordings. 2) Analyze the recorded cortical Current Source Density (CSD) using Principal Components Analysis (PCA). Model the CSD patterns expected from the matrix and core thalamocortical afferents using reconstructions of supragranular and infragranular pyramidal cells, their population distribution, and terminations of matrix and core afferents. Hypothesis: The main spatiotemporal CSD components contributing to the spindle identified with PCA will correspond to the CSD components modeled to result from activation of matrix and core thalamocortical afferents. 3) Construct neuronal models based on Hodgkin-Huxley ionic currents that include cortical cells, thalamic reticular nuclear cells, matrix and core thalamic relay cells, and their interconnections in a minicolumn. Hypothesis: The spatiotemporal patterns of currents predicted in different cortical layers will match those obtained using the methods described in modeling aim 2. 4) Scale up the cortical minicolumn network model using simplified neural models to a spatially accurate cortical model that can generate EEG/MEG patterns. Hypothesis: The predicted EEG and MEG from the model will match those observed in recordings, with the involvement of matrix vs core thalamocortical systems corresponding to the parameters derived in modeling aim 1. 5) Develop a statistical model that matches the properties of the scaled up model and analyze it with methods from statistical physics. Hypothesis: The matrix system controls the effective cortical connectivity and the core system controls the local correlation length. Collaborative research Recordings from humans will be performed at MGH (Cash), with analysis and modeling at UCSD (Halgren and Sejnowski). These 3 teams of PIs, students and postdoctoral fellows will interact on a daily basis during the research and will meet formally at least once a year to assess progress and plan new experiments.